Evidence of Silicene in Honeycomb Structures of Silicon on Ag(111)

Letter

pubs.acs.org/NanoLett

Evidence of Silicene in Honeycomb Structures of Silicon on Ag(111) † † † ‡ † † † † Baojie Feng, Zijing Ding, Sheng Meng, Yugui Yao, , Xiaoyue He, Peng Cheng, Lan Chen,*, † and Kehui Wu*, † Institute of Physics, Chinese Academy of Sciences, Beijing 100190, China ‡ School of Physics, Beijing Institute of Technology, Beijing 100081, China

ABSTRACT: In the search for evidence of silicene, a two- dimensional honeycomb lattice of silicon, it is important to obtain a complete picture for the evolution of Si structures on Ag(111), which is believed to be the most suitable substrate for growth of silicene so far. In this work we report the finding and evolution of several monolayer superstructures of silicon on Ag(111), depending on the coverage and temperature. Combined with first- principles calculations, the detailed structures of these phases have been illuminated. These structures were found to share common building blocks of silicon rings, and they evolve from a fragment of silicene to a complete monolayer silicene and multilayer silicene. Our results elucidate how silicene forms on Ag(111) surface and provides methods to synthesize high-quality and large- scale silicene. KEYWORDS: Silicene, Ag(111), scannning tunneling microscopy, molecular beam epitaxy, first-principles calculation

ith the development of the semiconductor industry of silicene and optimizing the preparation procedure for W toward a smaller scale, the rich quantum phenomena in growing high-quality silicene films, it is important to build a low-dimensional systems may lead to new concepts and complete understanding of the formation mechanism and ground-breaking applications. In the past decade graphene growth dynamics of possible silicon structures on Ag(111), has emerged as a low-dimensional system for both fundamental which is currently believed to be the best substrate for growing research and novel applications including electronic devices, silicene. 1−4 energy storage, and transparent protection layer. Inspired by In this Letter, we present a systematic study of the self- the fruitful results based on graphene, recently a lot of interest organized superstructures formed by submonolayer silicon 5−7 has been drawn to group IV (Si, Ge) analogs of graphene. It grown on Ag(111), by STM and scanning tunneling spectros- has been theoretically shown that silicene, with Si atoms packed copy (STS). We found that, depending on the substrate in a honeycomb lattice, like graphene, is a new massless Dirac temperature and silicon coverage, several monolayer super- Fermion system.5,8 Compared with that of graphene, the − structures can form on Ag(111). These superstructures are stronger spin orbit coupling in silicene may lead to a distinct from any known surface structures of bulk silicon and detectable quantum spin Hall effect (QSHE) and other 8−11 are characterized by honeycomb building blocks and structures. attractive properties. The compatibility of silicene with At sufficiently high temperature and Si coverage, monolayer silicon-based nanotechnology makes this material particularly and multilayer silicene films were grown. Combined with first- interesting for device applications. principles calculations, the structural models of these phases are As the theoretical studies on silicene are rapidly increasing, proposed, and their evolution with temperature and Si coverage the major challenge in this field is now the preparation of high- quality silicene films. However, to date, there is still no solid is discussed. Our work provides a complete understanding of evidence for the observation of a silicene film. There have been the structure evolution of Si on Ag(111), which is desirable for a few works on the formation of silicene nanoribbons on fabrication of high-quality silicene and exploring its novel Ag(110) with graphene-like electronic signature.12,13 The only physics and applications. published work on the preparation of silicene-like sheets was Experiments and Methods. Experiments were carried out reported by Lalmi et al. on Ag(111).14 They showed scanning in a home-built low-temperature STM with a base pressure of 5 × −11 tunneling microscopy (STM) images of a honeycomb 10 Torr. Single-crystal Ag(111) sample was cleaned by monolayer structure that resembles monolayer graphene cycles of argon ion sputtering and annealing. Silicon was ≈ structure. However, in their experiment the observed lattice evaporated from a heated wafer ( 1200 K) onto the preheated − constant was about 17% smaller than the theoretically proposed substrate. The deposition rate of silicon was kept at 0.08 0.1 model or the value of bulk silicon. Such a huge compression of the lattice is rather unlikely to be induced by the strain between Received: March 17, 2012 the film and the substrate. Their results therefore remain to be Revised: May 29, 2012 confirmed and understood. For the purpose of finding evidence Published: June 1, 2012

ML/min (here 1 monolayer refers to the atomic density of a ideal silicene sheet). The STS data were acquired using a lock- in amplifier by applying a small sinusoidal modulation to the tip bias voltage (typically 10 mV at 676 Hz). All our STM experiments were carried out at 77 K. First-principles calculations were performed within the framework of density functional theory (DFT) using projected augmented wave (PAW)15,16 pseudopotentials and the Perdew−Burke−Ernzerholf (PBE)17 form for exchange− correlation functional, as implemented in Vienna ab initio simulation package (VASP).18 During calculations, the structures were relaxed without any symmetry constraints using a plane-wave energy cutoff of 250 eV. The convergence of energy is set to 1.0 × 10−4 eV. The relaxation process continues until forces are below 0.01 eV/Å. Results and Discussions. Silicon atoms deposited on Ag(111) tend to form clusters or other disordered structures when the substrate temperature is below 400 K during growth (data not shown here). As substrate temperature increases to 420 K, two ordered phases form, as shown in Figure 1. The less ordered phase consists of close-packed protrusions (labeled T), and the highly ordered phase exhibits honeycomb structures (labeled H). The two phases can coexist on the surface within a large coverage range, from 0.5 to 0.9 ML (Figure 1a,b, respectively). The coexistence of phases H and T indicates that these two phases have quite similar formation energy and stability. One can therefore expect similarity and relations between the atomic structure of these two phases. Indeed, the high-resolution STM images in Figure 1c,d show that every big bright protrusion in both phase T and H is indeed composed of three smaller spots that we refer to as a “trimer”, although in phase H the trimers are perfectly ordered, while in phase T they exhibit some irregularity and distortion when looked at closely. The two Figure 1. (a) STM image (Vtip = 1.2 V) of 0.5 ML silicon atoms phases share the same periodicity of 1.18 nm, therefore the deposited on Ag(111) surface at substrate temperature of 420 K. The density of trimers in phase H is just twice of that in phase T. areas with phase T were marked by “T”, while the areas with phase H ff − We further performed STS measurements for the two phases (with di erent rotation angles) were marked by H1 H3, respectively. − (Figure 1e). The dI/dV curves exhibit very similar features of (b) STM image (Vtip = 1.5 V) of 0.9 ML silicon atoms deposited on electronic density of states (DOS), which strongly implies that Ag(111) surface at substrate temperature of 440 K. The areas of phase − the two phases may share some common building blocks in T and H are labeled. Notably H1 H3 mark areas with phase H in different orientations. (c,d) High-resolution STM images (8.5 × 8.5 their atomic structures. In fact, in Figure 1c, there is a 2 − noticeable point showing a corner hole and six protrusions nm , Vtip = 1.0 V) showing the atomic structure of phase T and H,  respectively. The blue rhombuses in (c,d) indicate the unit cells of two surrounding it a characteristic signature of formation of phase phases. (e) dI/dV spectra taken at areas of phase T (red) and H H. Moreover, we notice that phase T prefers to form at a lower (black), respectively. The curves have been shifted vertically for clarity. Si coverage and a slightly lower temperature as compared with phase H. With the increase of substrate temperature and coverage, phase T decreases in percentage and eventually of Si equals exactly four times the lattice constant of Ag. If one disappears completely at 460 K. Meanwhile, phase H can assume the observed H phase to be the theoretically proposed spread over the surface if the coverage is sufficiently high. This silicene, it is possible to obtain a 3 × 3 superstructure by placing means that phase H is more stable than phase T, and phase T the silicene lattice in parallel with the 1 × 1 Ag lattice. However, can be regarded as a precursor state of phase H. the crucial point in such moirépattern models is that the We now face a direct question whether these two phases, periodicity of the superstructure, or essentially the moiré especially the well-ordered H phase, are the theoretically pattern, is strictly linked with the relative orientations of the proposed silicene. The phase T can be excluded first due to the two overlapping lattices. If one obtain a 3 × 3 moirépattern in significantly lower density of Si than that of phase H. We notice one orientation, it will be impossible to observe the same that the periodicity of 1.18 nm is almost exactly four times the pattern in another inequivalent crystallographic orientation. lattice constant of Ag(111) 1 × 1 surface, 0.29 nm, or three However, as we show in Figure 1a, we have observed the times the lattice constant of silicene, 0.38 nm. Therefore both formation of 3 × 3 domains on the same Ag terrace, with phases H and T can be written as 4 × 4 reconstruction with different orientations which are obviously inequivalent. This respect to the 1 × 1 Ag substrate or 3 × 3 reconstruction with simple experimental fact excludes the possibility that the 3 × 3 respect to 1 × 1 silicene lattice (in this Letter we refer to as 3 × structure is a silicene lattice placed on 1 × 1 Ag. In another 3). In fact, Ag/Si system is known as a typical “magic words, the 3 × 3 reconstruction should come from the structure mismatched” system, such that three times the lattice constant of the overlayer itself, instead of from the commensuration

3508 dx.doi.org/10.1021/nl301047g | Nano Lett. 2012, 12, 3507−3511 Nano Letters Letter between the overlayer and the substrate. It is, however, noticeable that the 3 × 3 reconstruction is most clearly resolved in one major crystallographic orientation, while in other orientations, some irregular distortion of the lattice is seen, which should come from the inﬂuence of substrate Ag lattice. As noted above, the periodicity of 1.18 nm is three times the lattice constant of Si(111), 0.38 nm, which is also close to the calculated lattice constant of silicene.8 Based on STM observation of the characteristic corner hole structure, we propose a model of phase H as shown in Figure 2a. In this

Figure 2. (a,b) High-resolution STM images superposed with calculated model of phase H and T, respectively. The red and gray balls in the models represent buckled and unbuckled silicon atoms, respectively. The blue rhombus in (a) and (b)indicate unit cells as shown in Figure 1(c) and (d). The double arrows indicate that the rings can rotate randomly along their centers. model, the corner holes are due to missing of a hexagonal silicon rings in each 3 × 3 cell of a complete honeycomb silicene structure. This model has been confirmed by first- × 2 principles calculations. In the calculation the structure was Figure 3. (a) A derivative STM image (200 200 nm , Vtip = 1.43 V) modeled with low-buckled silicene lattice8 with missing silicon of 0.9 ML silicon atoms deposited on Ag(111) surface at substrate × 2 rings at the corners, and the six Si atoms around the corner are temperature of 480 K. (b) High-resolution STM image (15 15 nm , − ́ hydrogenated to saturate the Si dangling bond and simplify the Vtip = 1.0 V) showing the atomic structure of moire patterns. The calculations. After relaxation, there are no in-plane changes of bright areas exhibit complete honeycomb rings with a period of 1.0 nm, while other areas are defective and disordered. The angle between the position of silicon atoms, but the atoms close to the corner the orientation of the hexagonal rings and the direction of moiré holes (red atoms in Figure 2a) move upward, corresponding patterns is 30°. (c) dI/dV spectra taken at the moirépattern phase, in well with the trimer feature observed by STM. which a peak at 0.3 V and a shoulder at 0.9 V are observed. (d) Based on the atomic structure of phase H, the understanding Calculated model of √7 × √7 superstructure of silicene. The gray, of the atomic structure of phase T becomes straightforward. yellow, and red balls represent the silver, lower silicon, and higher Because the STM observation shows that the density of Si silicon atoms, respectively. (e,f) Experimental and simulated STM trimers in phase T is half of that in phase H, we construct the images (1.0 eV above Fermi energy) showing the similar structure model of phase T by removing half of the silicon rings in phase features and unit cell of lattice. H, leaving only one hexagonal silicon ring per 3 × 3 unit cell, as shown in Figure 2b. This model has also been validated by first- orientation of moirépattern is along the ⟨110̅⟩ direction of principles calculations. Similar to the calculation of phase H, we Ag(111), and the period is about 3.8 nm. The high-resolution chose low-buckled silicene rings, saturated by hydrogen atoms STM image in Figure 3 indicates that a few complete as in the original structure. The calculation results show that honeycomb rings with lattice period about 1.0 nm are observed this model is stable. Each trimer corresponds to a buckled at the bright part of the moirépattern, and the other parts are silicon ring with three Si atoms moving upward. Such a rather defective and disordered. Additionally the angle between structure can be considered as self-assembly of hexagonal the direction of moirépattern and honeycomb structure is silicon rings stabilized by weak van der Waals force and about 30°. The dI/dV spectra measured on this structure shows interaction between Si and Ag(111). Compared with the a peak at 0.3 V and a shoulder at 0.9 V, which is distinct from honeycomb arrangement of Si rings connected by covalent that of phase T and H and indicating an essentially different bonds in phase H, the weak connection of Si rings in phase T structure formed. might explain the observed more disordered trimer structure, as Considering that the complete honeycomb structure with 1.0 compared with the highly ordered trimer structure in phase H. nm periodicity is only observed at special positions on surface, When the substrate temperature during silicon growth we proposed that this honeycomb superstructure consists of reaches 480 K, the silicon structure exhibits another phase fragments of single layer of silicene with strong interaction with with obvious moirépattern, which is long-range ordered and the Ag(111) substrate. In order to clarify our supposition, first- can spread over the whole surface as shown in Figure 3a. The principle calculations have been performed. The structure

3509 dx.doi.org/10.1021/nl301047g | Nano Lett. 2012, 12, 3507−3511 Nano Letters Letter model is constructed as single layer, low-buckled silicene being in registry with five Ag(111) planes. Except for the two bottom Ag layers, all atoms are relaxed during the geometry optimization. The energetically stable structure is shown in Figure 3d. From the calculation results we find that silicon atoms directly above a silver atom (red balls in Figure 3d) are higher than other silicon atoms. As a result these atoms should be observed as bright protrusions in STM image and forming a √7 × √7 superstructure with respect to silicene or (2√3 × 2√3)R30° superstructure with respect to Ag(111). This gives a larger honeycomb lattice with period 0.386 × √7 = 1.02 nm, in accordance well with our experimental data. The simulated STM image according to the calculated model is shown in Figure 3f. The similar structure features and lattice period, as observed in experimental STM images (Figure 3e), strongly support our suggested model. Actually there is a slight deviation between the lattice constant of √7 × √7 (1.02 nm) superstructure of silicene from that of 2√3 × 2√3 (1.00 nm) lattice of Ag(111), which result in the formation of the moirépattern. The optimized structural model in Figure 3d shows the hexagonal rings of silicene are twisted due to the strong interaction between silicon atoms and silver substrate. The bright parts of moirépattern are where the positions of atoms in silicene are little deviated from that of Ag(111), which make the honeycomb superstructure stable enough to keep the hexagonal rings complete. In other parts of moirépattern, the larger deviation of position between atoms in silicene and those of Ag(111) lead to unstable honeycomb structure and eventually breaks the hexagonal rings of silicene, resulting in defective and disordered structures. The disordered structures were not obtained in our calculations because the unit cell we ́ choose is much smaller than that of a moire pattern. The angle × 2 √ × √ Figure 4. (a) 3D STM image (30 30 nm , Vtip = 1.0 V) of a single between the lattice direction of the 7 7 superstructure layer of silicene island across a step edge of Ag(111). (b) High- and ⟨110̅⟩ direction of Ag(111) is 30°, so the angle between the × 2 resolution STM image (8 8nm, Vtip = 1.2 V) of one monolayer direction of moirépattern and ⟨110̅⟩ direction of Ag(111) silicene terrace showing the √3 × √3 honeycomb superstructure with should be zero, which has been confirmed by our experiments. the period of 0.64 nm. (c) Top and side views of schematic model of As the substrate temperature reaches 500 K and the coverage √3 × √3 superstructure of silicene. The red, gray, and green balls is up to 0.8 ML, we observed dense honeycomb structure which represent the upper buckled, in plain, and lower buckled Si atoms, we identify as silicene. The STM image in Figure 4a shows a respectively. The √3 × √3 honeycomb superstructure is indicated by × 2 one-atom-thick silicene sheet across the step edges of the the black hexagon. (d) STM image (54 54 nm , Vtip = 1.5 V) of 1.2 Ag(111) surface without losing continuity of the atomic lattice, ML silicon atoms deposited on Ag(111) surface at substrate 19 temperature of 500 K showing the second layer of silicene formed which is similar to graphene grown on metal surfaces. The on the first layer of silicene. (e) dI/dV spectra taken on the first high-resolution STM image of Figure 4b shows a honeycomb ff × (black) and second (red) layers of silicene, respectively. (f) High- structure. However, di erent from the reported 1 1 structure resolution STM image (10.5 × 10.5 nm2, V = 1.5 V) of area as 14 tip of silicene, the lattice period of the honeycomb structure we marked by the white rectangle in (d) showing atomic structure of the observed is about 0.64 nm, which is corresponding to a √3 × first and second layers of silicene simultaneously. √3 honeycomb superstructure with respect to the 1 × 1 silicene lattice. For the general low-buckled silicene model,8 epitaxially grown on metals,21,22 we did not observe moiré three Si atoms in one hexagonal ring are upper buckled but still patterns in our film, which may originate from the weak exhibit 1 × 1 lattice (we named the AB configuration). To interaction between silicene and the metal substrate.23 A typical explain √3 × √3 superstructure, we give a symmetric-buckled dI/dV spectrum obtained at the silicene terrace (black curve in silicene model shown in Figure 4c. In this structure model, the Figure 4e) shows a shoulder at 0.3 V and a peak at 0.9 V, which six Si atoms in one hexagonal ring are not in plain: two atoms is similar as the LDOS distribution measured on moirépatterns are buckled upward (red atoms in Figure 4c) and one atom is phase. Another remarkable feature is a small dip located at 0.5 buckled downward (green atoms in Figure 4c), which we eV which is corresponding to the Dirac point (DP) of silicene. named the ABA̅configuration. The DFT calculations showed The dip is not very obvious compared with that of ABA̅configuration with the lattice period determined by our graphene,24,25 which is probably due to the pronounced experimental data are more stable than AB configuration (the electronic DOS of the underlying Ag(111) substrate super- detail of calculations is shown in ref 20). The upper buckled Si imposed on the dI/dV spectra. The deviation of the energy atoms are resolved by STM as the √3 × √3 honeycomb position of DP from the Fermi energy may stem from the superstructure. Based on this model and the lattice constant of charge transfer from the Ag(111) surface to silicene. silicene in our experiment, the atomic density of silicene is The assignment of the above phase as silicon gets a direct calculated as 1.69 × 1015 cm−2.Different from graphene proof by the observation of second layer of silicene at higher

3510 dx.doi.org/10.1021/nl301047g | Nano Lett. 2012, 12, 3507−3511 Nano Letters Letter coverage, as shown in Figure 4d. The high-resolution STM (7) Kara, A.; Leandri, C.; Davila, M. E.; Padova, P.; De; Ealet, B.; image of Figure 4f shows the atomic structure of the first and Oughaddou, H.; Aufray, B.; Lay, G. Le J. Supercond. Novel Magn. 2009, − second layers silicene simultaneously. It is obvious that the 22, 259 263. second layer silicene also exhibits a √3 × √3 honeycomb (8) Liu, C. C.; Feng, W. X.; Yao, Y. G. Phys. Rev. Lett. 2011, 107, superstructure, which indicates that the √3 × √3 honeycomb 076802. (9) Liu, C. C.; Jiang, H.; Yao, Y. G. Phys. Rev. B 2011, 84, 195430. superstructure should originate from free-standing silicene and fl (10) Kane, C. L.; Mele, E. J. Phys. Rev. Lett. 2005, 95, 226801. is not in uenced by the Ag(111) surface. This layer-stacked (11) Bernevig, B. A.; Hughes, T. L.; Zhang, S. C. Science 2006, 314, silicon structure is similar to graphite and is a new structural 1757−1761. phase of silicon, which may host many novel properties. The (12) Aufray, B.; Kara, A.; Vizzini, S.; Oughaddou, H.; Leandri,́ C.; dI/dV spectrum on second layer of silicene (red curve in Figure Ealet, B.; Lay, G. Le Appl. Phys. Lett. 2010, 96, 183102. 4e) resembles that on the first layer. This striking similarity (13) Padova, P. De; et al. Appl. Phys. Lett. 2010, 96, 261905. between the LDOS of monolayer and the bilayer silicene can (14) Lalmi, B.; Oughaddou, H.; Enriquez, H.; Kara, A.; Vizzini, S.; also confirm our predation that interactions between the Ealet, B.; Aufray, B. Appl. Phys. Lett. 2010, 97, 223109. (15) Vanderbilt, D. Phys. Rev. B 1990, 41, 7892. monolayer silicene and Ag(111) are as weak as that between ̈ the two silicene layers. (16) Blochl, P. E. Phys. Rev. B 1994, 50, 17953. (17) Perdew, J. P.; Burke, K.; Ernzerhof, M. Phys. Rev. Lett. 1996, 77, Even at a substrate temperature of about 500 K during ́ 3865. growth, silicon atoms tend to form the moire pattern phase if (18) Kresse, G.; Hafner, J. Phys. Rev. B 1993, 47, 558. the coverage of silicon is considerably less than 0.8 ML. This (19) Coraux, J.; N’Diaye, A. T.; Busse, C.; Michely, T. Nano Lett. indicates that the atomic density of silicene is higher than that 2008, 8, 565−570. of the moirépattern phase, which justifies our structural model (20) Chen, L.; Liu, C. C.; Feng, B. J.; He, X. Y.; Cheng, P.; Ding, Z. again. Increasing the substrate temperature to above 600 K, no J.; Meng, S.; Yao, Y. G.; Wu, K. H.; arXiv:1204, 2642. structure of silicon can be observed anymore, leaving only a (21) Marchini, S.; Günther, S.; Wintterlin, J. Phys. Rev. B 2007, 76, bare Ag(111) surface. Furthermore, if the sample of silicene on 075429. fi (22) Land, T. A.; Michely, T.; Behm, R. J.; Hemminger, J. C.; Comsa, Ag(111) is annealed up to 600 K, silicene lm will also − disappear. The upper temperature limit that our silicene film G. Surf. Sci. 1992, 264, 261 270. 22,26,27 (23) Sutter, E.; Acharya, D. P.; Sadowski, J. T.; Sutter, P. Appl. Phys. can endure is considerably lower that that of graphene. Lett. 2009, 94, 133101. This might stem from the weaker interaction between silicene (24) Zhang, Y. B.; Brar, V. W.; Wang, F.; Girit, C.; Yayon, Y.; and Ag(111) substrate, as compared with that of graphene on Panlasigui, M.; Zettl, A.; Crommie, M. F. Nat. Phys. 2008, 4, 627−630. Ir(111) surface. (25) Zhao, L.; et al. Science 2011, 333, 999−1003. Conclusion. We have systematically investigated the (26) Sutter, P.; Flege, J.; Sutter, E. Nat. Mater. 2008, 7, 406−411. structure evolution of silicene on Ag(111). With the increase (27) Yu, Q.; Lian, J.; Siriponglert, S.; Li, H.; Chen, Y. P.; Pei, S. Appl. of the substrate temperature, silicon atoms on Ag(111) Phys. Lett. 2008, 93, 113103. overcome potential barriers and form some metastable structural phases, such as self-assembled honeycomb building blocks (phase T and H) and incomplete silicene film (moiré pattern structure). The most stable phase, decoupled monolayer and bilayer silicene film, will appear eventually. This work provides methods to fabricate high-quality silicene, which is essential to investigate its novel properties, and brings it closer to the use in nanotechnology and other related areas.

■ AUTHOR INFORMATION Corresponding Author *E-mail: [email protected]; [email protected] Notes The authors declare no competing ﬁnancial interest.

■ ACKNOWLEDGMENTS This work was supported by the NSF of China (Grant nos. 11074289 and 91121003) and the MOST of China (Grant nos. 2012CB921703 and 2009CB929101).

■ REFERENCES (1) Geim, A. K.; Novoselov, K. S. Nat. Mater. 2007, 6, 183−191. (2) Castro Neto, A. H.; Guinea, F.; Peres, N. M. R.; Novoselov, K. S.; Geim, A. K. Rev. Mod. Phys. 2009, 81, 109−162. (3) Pumera, M. Chem. Rec. 2009, 9, 211−223. (4) Kim, K. S.; et al. Nature 2009, 457, 706−710. (5) Cahangirov, S.; Topsakal, M.; Akturk, E.; Sahin, H.; Ciraci, S. Phys. Rev. Lett. 2009, 102, 236804. (6) Kara, A.; Enriquez, H; Seitsonen, A. P.; Lew Yan Voon, L. C.; Vizzini, S.; Aufray, B.; Oughaddou, H. Surf. Sci. Rep. 2012, 67,1−18.

3511 dx.doi.org/10.1021/nl301047g | Nano Lett. 2012, 12, 3507−3511